Mean-Variance Optimisation: The Engine Behind Every Portfolio

Ask most people how to build an investment portfolio and they will say something about picking good stocks or spreading across different things. But professional portfolio construction goes far deeper. It starts with a precise question: for any given level of risk, what combination of assets produces the highest expected return? The answer comes from mean-variance optimisation, the most widely used framework in the world for setting asset allocation policy.

What the Framework Actually Does

The core idea in plain English: Every asset class has an expected return (the reward) and volatility (how much that return bounces around year to year). MVO finds the single best mix of assets: the highest possible return for a chosen level of volatility, or the lowest volatility for a chosen expected return. It does this for every possible risk level, tracing a line called the efficient frontier.

Mean-variance optimisation (MVO): A mathematical process that picks the best mix of asset classes by trading off expected return against variance. Developed by Harry Markowitz in 1952

Efficient frontier: The curve showing all optimal portfolios. Every point on it offers the highest possible return for its level of risk. Portfolios below the curve are suboptimal

Variance: A statistical measure of how much returns fluctuate. High variance means the return in any given year is hard to predict

The framework needs three inputs for each asset class: its expected return, its expected volatility (standard deviation), and its correlation with every other asset class. Correlations matter because the whole point of diversification is combining assets that do not all move together. When two assets have a correlation below 1, combining them produces a portfolio whose risk is less than the weighted average of their individual risks.

The optimiser also needs to know how risk-averse the investor is. This is captured by a number called the risk aversion coefficient, usually denoted lambda. A low coefficient (around 2) means the investor tolerates volatility willingly in pursuit of higher returns. A high coefficient (around 6) means they will accept lower returns to keep volatility controlled. Most real investors fall between 1 and 10, with a coefficient of 4 representing a moderately cautious person.

How the Efficient Frontier Works in Practice

The efficient frontier is a curve, not a single portfolio. At the far left sits the global minimum variance portfolio: the combination with the lowest possible volatility. At the far right sits the maximum return portfolio: typically a 100% allocation to the single highest-returning asset class. Between these extremes, every point on the frontier is efficient because no other combination delivers more return for that level of risk.

As you move from left to right along the frontier, each additional unit of risk bought progressively less additional return. The frontier is steep on the left (small risk additions deliver meaningful return gains) and flat on the right (large risk additions deliver little extra return). This is why very aggressive portfolios often disappoint relative to their risk: the frontier's slope has flattened to almost nothing.

The investor's risk aversion coefficient identifies the specific point on the frontier that maximises their utility. This is the portfolio where the trade-off between extra return and extra risk exactly matches the investor's preference.

The Economic Balance Sheet: Seeing the Full Picture

Why standard MVO misses something important: Standard MVO only looks at the financial portfolio. But a working person's biggest asset is often not in their brokerage account. It is their human capital, the present value of future earnings from their job. A family home is another major non-financial asset. Ignoring these hidden assets leads to portfolios that look well-diversified on paper but are not in reality.

Human capital: The present value of all future earnings from employment. For a young professional, this can dwarf their financial savings

Economic balance sheet: An extended personal balance sheet that includes human capital, real estate, and other non-financial assets alongside the investment portfolio

The nature of human capital matters as much as its size. A tenured university professor has very stable, inflation-linked income. Their human capital behaves like a large, long-duration inflation-linked bond. A commission-based salesperson has volatile income that rises and falls with the economy. Their human capital behaves more like equity. Two people with identical financial portfolios need very different investment approaches once their human capital is considered.

The Global Market Portfolio: A Principled Starting Point

Global market portfolio: A theoretical portfolio holding every investable asset in the world in proportion to its market capitalisation. It represents the collective judgment of all investors worldwide

One of the hardest inputs for MVO is expected returns. Analysts frequently get these wrong. Small errors in expected returns can produce dramatically different and unintuitive portfolio weights. One disciplined solution is to anchor return estimates to the global market portfolio using a technique called reverse optimisation.

Reverse optimisation works backwards. Instead of assuming returns and calculating optimal weights, it assumes the market-capitalisation weights are optimal and works backwards to calculate what expected returns must be if that is true. The result: internally consistent return estimates that reflect the collective wisdom of all market participants.

The global market portfolio also provides a reference point for monitoring. If an investor's recommended allocation deviates substantially from global market weights in an asset class, there should be a specific, defensible reason. Simply preferring a particular market without a clear rationale is not enough.